Hi, I'm

Abhigyan Tiwari

Software, Systems & AI Engineering

Building distributed platforms, agentic AI systems, and high-performance backend infrastructure.

View Work GitHub

Not just a Developer.

I write high-performance code.

01. About Me

I am a Computer Science and Engineering undergraduate at NIT Silchar, deeply passionate about solving complex engineering problems at scale.

My focus lies at the intersection of Distributed Systems, Backend Architecture, and Agentic AI. I love building systems that are robust, highly concurrent, and deeply integrated with modern cloud infrastructure.

Languages & Databases

C++, Python, Go, MySQL

Frameworks & Cloud

FastAPI, React, WebSockets, Docker, AWS (Lambda, DynamoDB, EC2), Git

AI, ML & Systems

Agentic AI, RAG, Pandas, Polars, gVisor

Honors & Awards

  • ⭐ Semi-finalist at GridLock Hackathon (Flipkart & Bengaluru Traffic Police)
  • 🚀 Among Top 2.1% teams (Finalist) in IICPC Summer Trading Hackathon 2026
  • 🏆 2nd Prize at Arbitrage Arena 2026 (National-level Quant Finance, IISc)
  • 🥉 3rd Position, IISc Honour Code (National Level Hackathon, Feb 2025)
  • ⭐ Semi-finalist at Tata Innovent 2026
  • 💻 Achieved 2,245th rank in Round 1 of Meta HackerCup 2025

02. Featured Projects

Fintech & Agentic AI

Pratyānayana

Autonomous Merchant Revenue Recovery AI Agent. Predicts payment recoverability and executes bounded Retry/Wait/Switch/Stop decisions using causal attribution and expected-value economics.

  • Event-driven Razorpay recovery with idempotent webhooks.
  • Benchmarked 5,000 journeys recovering 3.65 Cr with 45.5L incremental lift.
  • FastAPI
  • CatBoost
  • LLM
GitHub ↗

Distributed Systems

Vahini

A distributed benchmarking platform on AWS for automated correctness validation and live scoring of trading engines.

  • Secure execution pipeline using Docker and gVisor to sandbox untrusted participant code.
  • Worker-coordinator architecture sustaining 1.3K+ TPS under 1,000 concurrent traders.
  • Go
  • Docker
  • AWS EC2
GitHub ↗

Serverless & LLM

After Loss

A rules-driven platform to discover and claim a deceased person's financial assets from documents.

  • Automated asset routing via AWS Textract, Step Functions, and Lambda.
  • Cognito/Cedar authorization, PII masking, and LLM-based cited assistance (27/27 citations).
  • DynamoDB
  • React
  • Cedar
GitHub ↗

Developer Tools

DockAgent

An intelligent, agentic CLI tool written in Go that autonomously generates and self-heals Dockerfiles by hooking into the local Docker daemon and feeding build errors back to the LLM.

  • Go
  • Docker API
  • Gemini AI
GitHub ↗

03. Open Source

OpenCode IIITA

Contributed 31 merged PRs across 13+ repositories. Implemented algorithmic solutions (Binary Search/Prefix sums), performed dataset EDA with imputation, and solved competitive coding challenges.

CSS NITS

Contributed to ComplaintBox_CPP by eradicating database vulnerabilities to protect the database, optimizing compile times by removing redundant headers, and improving the CLI menu layout.

Apothesis

Resolved critical cross-platform build failures on Windows. Fixed CMake toolchain compatibility and enforced C++17 requirements. Merged upstream into the core repository.

Layer5 & Meshery

Open source contributor to Layer5 / Meshery (Cloud Native Management Plane). Logged contributions to the community and ecosystem.